The interaction volume is typically smaller than capture range to leave a margin for tracking jitter

What it is

The interaction volume is conventionally drawn a ring inside capture range, not to look conservative, but to leave a margin for tracking jitter. On the lip of the capture envelope, joints appear and vanish, depth noise grows, and a small shake throws the hand out of frame. If valid input is opened all the way to the line where hardware can just still see, work along the boundary flickers between “counts” and “tracking just died.” A smaller interaction volume keeps boundary work inside the stable tracking band.

Why it happens

Depth and keypoint error rise with angle off the optical axis and with distance to either end of the range: fewer edge pixels, uneven IR, stereo matching failing on slanted surfaces. Ordinary hand tremor and breathing sway push a point that is nominally “just inside” back out. A hard containment test treats every exit as a loss. Shrinking the interaction volume by several centimetres to a decimetre or so is a policy decision to discard the already-unreliable rim of signal, trading that rim for continuity. The width of the margin should come from measured jitter of that device at the target distance, not from copying the slogan “the interaction volume should be smaller than capture range.” That sentence names a direction, not a distance.

Studying it

Along the six faces of the capture envelope—top, bottom, left, right, near, far—measure spatial standard deviation of joints and gap lengths of loss. Place the interaction-volume boundary at different percentiles (for example, 95% of jitter still inside) and have participants scrub or drag along the boundary. Dependent measures are dropouts, completion time, and reports of a sudden release. Compare zero margin (interaction volume = capture envelope) with a margin condition. This is edge stability, not “how far should a person stand”—stance comfort is a different question.

Where it stops holding

When capture range is already tiny (a near-field desktop module, a narrow phone selfie frustum), shrinking another layer leaves nowhere to put a hand; margin and operable space crowd each other. Full-body room-scale tracking jitters differently from near-field hands; a centimetre-scale margin may be too little or too much. Jitter at a multi-camera seam is not an outer lip; putting margin only on the outermost envelope misses interior cracks. If the interaction volume follows the body, the margin has to live in the relative frame; a leftover band nailed to the room wall does nothing.

Applying it

• Measure edge jitter on the target device at install distance, set the interaction-volume boundary inside the isosurface where continuous work stays stable, and write down the margin in centimetres.
• Do not grow the interaction volume out to the capture envelope just to fit one more standing person; the extra reach will be eaten by edge flicker.
• For regression, drag along each of the six faces: a mild “near the boundary” cue is allowed; tracking that repeatedly dies while still nominally inside the volume is not. Death means the margin is too thin, not that the user stood badly.

Related

• Same group: C4.20.1 Capture range is the volume in which the sensor can physically detect motion · C4.20.2 The interaction volume is the subspace of capture range treated as valid input · C4.20.3 Motion inside capture range but outside the interaction volume is easily mistaken for system failure
• Adjacent: C4.10 Sensing volume and interaction distance · C4.22 Perceiving interaction-volume boundaries
• Search: tracking jitter · margin · capture envelope